Power equipment serves as a critical component ensuring the stable operation of power systems, undertaking the task of electricity distribution. However, changes in the external environment may impact its normal functioning. Based on this, this paper proposes a risk level assessment model for intelligent water immersion alerts targeting power equipment. First, ArcGIS geographic information systems are employed to collect high-resolution LiDAR data from substations, thereby constructing digital twin data suitable for mirror analysis. Second, building upon the Stacking model within machine classification algorithms, a grid security situation early warning module is constructed. This module integrates geographic, meteorological data, and power equipment information as inputs, primarily serving to predict and warn of water immersion depths for power equipment. Finally, a case study analysis is conducted using a substation in a coastal city as the research subject. Results demonstrate that this model achieves significantly lower prediction errors for water immersion depth compared to benchmark models, with a recall rate of 99.98% for power outage warnings. Dynamic early warnings based on this framework enable power grid enterprises to proactively implement protective measures, mitigating the impact of external environmental changes on the stable operation of power systems.
